Jul 1, 2013

When Wholes Are Less Than Sums of the Parts

Headline national two party preferred results are clouding the dynamics and reality of what is happening at the state level - where elections are won and lost

Possum Comitatus — Editor of Pollytics

Possum Comitatus

Editor of Pollytics

Now that politics has reset itself, it’s probably worth looking at the polls since the change of PM – the results of which can pretty much be described as consensus

This is all very interesting and whatnot – Rudd poll boost, ALP becomes competitive etc etc…. see your nearest newspaper/blog for details.

But there’s something else worth talking about, and that’s what is going on underneath these topline national results.

Unless one side starts running away with the polling, or something weird happens in Qld like Rudd getting caught wearing a Blues jersey at the Origin decider, the headline two party preferred results may be a little misleading going in to the election – not in terms of them being wrong, but in terms of the seats that they would ordinarily deliver to each side. To understand why, we first need to look at what’s happening with Federal voting intention at the State level.

Both the Morgan SMS and ReachTEL polls had large sample sizes (5548 combined) and fortuitously we have the state breakdowns for each, allowing us to get a decent handle on the dynamics playing out at the State level. We also have an additional poll for Qld of 980 odd respondents which is not quite an internal poll, but it’s not quite not one either. It says the same things as the Qld components of ReachTEL and Morgan SMS (e.g ReachTEL has the ALP primary in Qld at 39.6, Morgan has it at 44 and the third party poll has the ALP primary sitting between them), so it’s all a much of a muchness – effectively just reducing our margin of error on the Qld numbers.

The “Swing to the ALP” figure is the change to the ALP two party preferred vote since the 2010 election. As we can see, there’s big movements to the ALP in Qld, but large movements away from them in Victoria and South Australia, with NSW and WA remaining static.

If we plug those numbers into Antony Greens Election Calculator, we get the ALP currently sitting on 77 seats, the Coalition 71 and 2 Independents (Katter and Wilkie). The Tasmania, ACT and NT results all come from small samples in either the ReachTEL or Morgan SMS results – so they’re a bit iffy, but you get the general picture.

My election simulation produces a similar result 76 seats to the ALP vs. 72 to the Coalition, with 2 Independents.

To understand how the ALP having a two party preferred national result of 48 or 49 can lead to getting a majority of seats, we need to look at what margins seats are sitting on in what states, and how swings in states produce wildly different seat gains/losses because of margin clustering.

The chart above shows the “swing to the ALP” half of the equation for LNP held seats on a 7% margin or less, or for seats requiring less than an ALP 60/40 average result in any state

Where each line crosses the bottom axis reflects what the ALP two party preferred result was at the 2010 election in each state – so WA (the black line) starts at 43.6%, Qld (the red line) starts at 44.9% etc. The points on each line represent a seat. As the ALP vote increases in each state, it shows roughly how many seats would be expected to fall to the ALP for a given change in two party preferred.

As you can see, with Qld coming off a relatively low base at the 2010 election, a 5% swing to the ALP would deliver around 9 seats to Labor in Qld simply because so many LNP held seats sit on relatively small margins. If the same swing occurred in WA it would only deliver 3 seats, in NSW 5 seats.

So  not only will Qld deliver more seats for a given swing than other states, we would also expect the swing in Qld to be larger than other states, simply because Rudd is PM (combined with the effect that his 2010 removal had on killing the ALP vote in Qld at the 2010 election)

Now let’s look at the other half of the equation – the swing away from the ALP

The big dangers here for Labor are NSW for any level of swing against them, and Victoria for relatively small swings.

In NSW, swings away from the ALP will catch seats on a 0.9% margin, then 1%, then 1.1%, then 1.5%, 2.7%, two on 4.2% and others on 4.4%, 5.1% and 5.2%

A 1.7% swing against the ALP in Victoria delivers 3 seats. But the next seat doesn’t fall on the pendulum after that until a swing of 5.8% against the ALP occurs.

So if the headline two party preferred results have underneath them big swings towards the ALP in Qld and relatively modest swings away from them in either NSW or Victoria (but not both), then we would expect to see the ALP win more seats than the national two party preferred might ordinarily suggest (for example, what we see now with the current polling).

This also tells is a little about the ALP’s path to 76 seats.

  1. Take Qld
  2. Either hold ground in both Victoria and NSW, or hold as much ground as possible in Victoria while aiming to make a few gains in NSW to offset probable losses like Dobell
  3. Take Qld

The major parties will start saying that the election will be won in Qld and NSW. For the ALP it will be taking Qld and holding NSW, for the Coalition it will be taking NSW and holding Qld. Yet considering that Qld is now much easier for the ALP, if Qld falls it goes a long way to balancing out any ALP losses in NSW and Victoria. If Qld falls one way and NSW the other, then things suddenly get really complicated.

Unless  someone breaks away in the polling.

But until or if they do, just keep an eye on the state breakdowns, because they’ll tell a more interesting and accurate story than the headline two party preferred figures.


(Visited 15 times, 1 visits today)


Leave a comment

14 thoughts on “When Wholes Are Less Than Sums of the Parts

  1. my say

    o see no up dates o well worth a look

  2. my say

    well left e and u and Andrew have become very pessimistic

    so I am her don’t know what they like to do here leave links or any thing but will see

  3. my say

    hi all thought I would take a look here cannot take the negativity of the poll bludger any more

    never new this was here

    so poss did u include that way old poll from morgan in YOUR graff please

  4. lefty e

    Hang on, this was 5 weeks ago. hah!

  5. lefty e

    Oops, that wasn meat for PB. But I assume youre flattered anyway Poss.

  6. lefty e

    Pessimists TAKE NOTE: polls can and will shift, but if the election was held *today*, modelling suggests the ALP would WIN IT, if swings are examined state-by-state, and taken to be uniform etc.

    And so says the mighty Possum:


    Ok, sure we need to know about TAS, but ITS ON.

    So I dont wanna hear any %$#*n defeatist whining ok? Get to it.

    COME ON! *Lleyton head-point*

  7. NathanA

    I guess the question I’m asking myself is how important will preferences be in the election? Are the polls likely to reflect the way preferences will be distributed, or will the preference flow from Katter’s party be a key factor in the election, even more so because he will be expected to poll much higher in qld? Or maybe a better way to ask the question is, how high does the lnp primary vote have to be for preferences not to matter?

  8. BGration

    State-by-state analysis makes a lot of sense and gets us down to counting seats. Good. But BETTER if you used electoral pendulums (pendula?). If you want graphic representation of swing trends and seats won or lost, the pendulum easily bests the graph.

  9. Socrates


    Thanks, that is very interesting and very encouraging for those who prefer their PMs not to be religious zealots.

    Regarding the headline number of seats in your calculation (ALP 77, Indie 2) does that include the effect of some individual seat changes? Realistically I think Labor is highly likely to lose Thomson’s seat in NSW, and Oakeshot and Windsor’s seats will go to the Nationals. Is that factored into the 77? If so, great 🙂

    You may get the Nate Silver award out of this one.

  10. Simon

    You probably should have made explicit that your prediction also involves Labor winning back Melbourne, and that without this outcome they’re back at 75.
    So that would be the fourth point on the ALP’s pathway to 76.

  11. morphy richards toaster

    Great, thorough analysis. A pity the mainstream media deems such in depth analysis too challenging for its target market.

    The question in my mind right now is, will people really vote this way, or are they just flirting with the polls?

  12. Disasterboy

    Thanks Pollytics. Its great to see a substantial analysis which clarifies what I was wondering. I’m cuious about “Others” at the moment. Psephologically and as Greens person. With HoR close that probably means the next Senate election will result in teh Greens with balance of power, even if none were relected. As is common leading to an election (vague unsubstantiated recollection) the Others figure tends do grow somewhet. WHere do they go and where might they go this time? With the new entries of pe3rsonality large parties Katter and Clive Plamer’s lot. do they have any chances with the odds and ends. FF and the ACP Sex Party and so on? Gee theres the Wikipedia party and Pirates too! Do any of them have a chance and if so, in which states? I think you did a bit of “Others” analysis a while back, but maybe it needs an update for current developments. Some pollsters have been doing bits of second tier party break downs. Is there more information they collect that we don’t usually see?

  13. canberra boy

    [sings]He’s back, he’s back, the possum is back… [/sings] [Does jig] [/jig]
    Long time no comment on national politics, possum. I was wondering what it would take to get you going again. Anyway, I hope you can spare the time to do some analysis and blog updates in response to the polling in coming weeks.

  14. ruawake

    Ran into Bill Gissane (ALP candidate for Fisher) today. Is very upbeat about beating Brough on everyone’s preferences. 🙂

Share this article with a friend

Just fill out the fields below and we'll send your friend a link to this article along with a message from you.

Your details

Your friend's details